Like Most People Neural Network Struggles to Write Opening Line

dejkrcbh

first_imgLet us know what you like about Geek by taking our survey. MIT’s AI Knitting System Designs, Creates Woven GarmentsResearchers Train AI To Feel Emotion, Too Stay on targetcenter_img It’s National Novel Writing Month (NaNoWriMo, to those in the biz). And who better to try their hand at penning a novel than a neural network?The annual creative writing event—the goal of which is to create a 50,000-word novel between Nov. 1 and 30—is for “anyone who has ever thought about writing a novel,” according to the website.Anyone, or anything, in this case.“When you’re faced with a blank page, sometimes it’s just hard to get started,” research scientist Janelle Shane wrote in a blog post. “I wanted to see if I could train a computer program to help.”Turns out, she couldn’t.To churn out anything remotely useful, the algorithm requires thousands of examples. Unfortunately, Shane could only find “a couple hundred” of the most famous opening lines.Malnourished, the neural network rebelled, basically reading back word-for-word what was fed into it.“Think of it like cramming for a test by memorizing instead of learning how to apply rules to solve problems,” Shane said.The result, while predictably hilarious, isn’t much help to aspiring writers, who could be slapped with a plagiarism suit if they tried passing off one of the computer’s lines as their own.Like this jumbled mess, appropriated from Charles Dickens’ A Tale of Two Cities:“The snow in the story of the best of times, it was the season of Darkness, it was the season of Light, it was the epoch of belief, it was the worst of times, it was the season of Light, it was the season of Darkness, it was the season of Light, it was the season of Light, it was the season of Darkness, it was the season of exploding past of Eller, and Junner, a long sunset side of the World.”(Honestly, use that rambling sentence in a publication, and you may have bigger problems than literary theft.)According to Shane, much of the computer’s output didn’t make sense, or was an “obvious mishmash” of famous lines. A few usable (if not thought-provoking) sentences, however, managed to slip in:“It was a wrong number that struggled against the darkness.”“The moon turned out to see me.”“The sky above the present century had reached the snapping point.”“I was born in the darkness.”(The algorithm obviously has an affinity for S. E. Hinton.)“Clearly, the neural network needed help,” Shane lamented. “Where could I get it more data?”Probably not the Bulwer-Lytton Fiction Contest, which invites people to write the worst first sentence.“I added them all,” she admitted of the 900-strong archive. “It didn’t help.”(Personally, I’d read a book that starts with “It was a dark and stormy night and the secret being a silver-backed gorilla.”)So, Shane put out a call for help: Following her successful crowdsourcing of Halloween costumes and D&D character backstories (coming soon), she is asking folks to share their own first line, that of their favorite story, or one from “every novel on your bookshelf.”Enter online as many as you like; Shane hopes to collect enough sentences by the end of the month to “give this another try.”last_img

Leave a Reply

Your email address will not be published. Required fields are marked *